No, I don't plan on writing everything. The goal of this thread is to figure out what existing software that may be used as building blocks.

But 2-4 weeks for a pre-study and then 3-4 months development time (excluding testing and documentation) is acceptable if the correct tools can be found.

I have done conceptual tests with a minimalistic byte-code interpreter just to verify that the target hardware can run multiple virtual machines, access data and produce meaningful results. But the language used was limited to only 32-bit integers - no structures or arrays. Writing a full front-end for C is totally out of the question. Writing a front-end for Pascal is possible.

I have at one time written a Mega88 emulator. I spent about a week for emulating the full instruction set, and one more week for timer, UART, RTC and SPI. I did that as a hobby project just because :)

But the wxWidgets GUI can be quite easily stripped, so modifying the code to run on the target hardware is not a big operation - there are about 100 lines of x86 assembler that would need to be replaced with C or ARM assembler. Modifying something like AVaRICE or avr-gdb to allow stimuli from a text file or from a serial link to the real hardware should not be a really big problem either.

Most probably, the AVR solution can be implemented to alpha state in less than a month. But it requires a closer look at licenses etc. But there are several other routes possible too.